samba - Dec 2017 - Samba does not report I/O error to the client

Dear Samba community,

I am facing the following problem. I have the following share in my
/etc/samba/smb.conf:

[external]
   comment = "External Hard Drive"
   path = /mnt/external
   read only = no

Sometimes it happens that external hard drive fails. Kernel issues USB reset and
the device /dev/sde1 in example below becomes invalid/inaccessible:

Dec 26 21:41:03.523585 kernel: usb 5-2: reset high-speed USB device number 3
using ehci-pci
Dec 26 21:41:03.979588 kernel: usb 5-2.1: new high-speed USB device number 100
using ehci-pci
Dec 26 21:41:04.167542 kernel: usb 5-2.1: New USB device found, idVendor=1058,
idProduct=1104
Dec 26 21:41:04.168074 kernel: usb 5-2.1: New USB device strings: Mfr=1,
Product=2, SerialNumber=3
Dec 26 21:41:04.168379 kernel: usb 5-2.1: Product: My Book
Dec 26 21:41:04.168687 kernel: usb 5-2.1: Manufacturer: Western Digital
Dec 26 21:41:04.168995 kernel: usb 5-2.1: SerialNumber: 575532553130303530353538
Dec 26 21:41:04.191879 kernel: usb-storage 5-2.1:1.0: USB Mass Storage device
detected
Dec 26 21:41:13.365071 kernel: blk_partition_remap: fail for partition 1
Dec 26 21:41:13.365382 kernel: EXT4-fs warning (device sde1): dx_probe:744:
inode #35151873: lblock 0: comm smbd: error -5 reading directory block
Dec 26 21:41:13.365522 kernel: blk_partition_remap: fail for partition 1
Dec 26 21:41:13.365658 kernel: EXT4-fs warning (device sde1): dx_probe:744:
inode #35151873: lblock 0: comm smbd: error -5 reading directory block

Filesystem correctly reports this situation:

$ ll /mnt/external
ls: reading directory '/mnt/external': Input/output error
total 0

However when I browse this share via Samba client, it reports that the directory
is simply empty:

$ smbclient //server/external -U dmitry%password -c ls
Domain=[WORKGROUP] OS=[Windows 6.1] Server=[Samba 4.5.12-Debian]
  .                                   D        0  Mon Dec 25 22:39:29 2017
  ..                                  D        0  Sun Apr 24 19:50:32 2016
                1922727728 blocks of size 1024. 45707296 blocks available

How can I change the behaviour so that I/O error is appropriately reported, e.g.
the client gets "Error reading directory"?

Many thanks.

-- 
With best regards,
Dmitry

On Wed, Dec 27, 2017 at 12:12:33AM +0100, Dmitry Katsubo via samba
wrote:
> Dear Samba community,
> 
> I am facing the following problem. I have the following share in my
> /etc/samba/smb.conf:
> 
> [external]
>    comment = "External Hard Drive"
>    path = /mnt/external
>    read only = no
> 
> Sometimes it happens that external hard drive fails. Kernel issues USB
reset and
> the device /dev/sde1 in example below becomes invalid/inaccessible:
> 
> Dec 26 21:41:03.523585 kernel: usb 5-2: reset high-speed USB device number
3 using ehci-pci
> Dec 26 21:41:03.979588 kernel: usb 5-2.1: new high-speed USB device number
100 using ehci-pci
> Dec 26 21:41:04.167542 kernel: usb 5-2.1: New USB device found,
idVendor=1058, idProduct=1104
> Dec 26 21:41:04.168074 kernel: usb 5-2.1: New USB device strings: Mfr=1,
Product=2, SerialNumber=3
> Dec 26 21:41:04.168379 kernel: usb 5-2.1: Product: My Book
> Dec 26 21:41:04.168687 kernel: usb 5-2.1: Manufacturer: Western Digital
> Dec 26 21:41:04.168995 kernel: usb 5-2.1: SerialNumber:
575532553130303530353538
> Dec 26 21:41:04.191879 kernel: usb-storage 5-2.1:1.0: USB Mass Storage
device detected
> Dec 26 21:41:13.365071 kernel: blk_partition_remap: fail for partition 1
> Dec 26 21:41:13.365382 kernel: EXT4-fs warning (device sde1): dx_probe:744:
inode #35151873: lblock 0: comm smbd: error -5 reading directory block
> Dec 26 21:41:13.365522 kernel: blk_partition_remap: fail for partition 1
> Dec 26 21:41:13.365658 kernel: EXT4-fs warning (device sde1): dx_probe:744:
inode #35151873: lblock 0: comm smbd: error -5 reading directory block
> 
> Filesystem correctly reports this situation:
> 
> $ ll /mnt/external
> ls: reading directory '/mnt/external': Input/output error
> total 0
> 
> However when I browse this share via Samba client, it reports that the
directory
> is simply empty:
> 
> $ smbclient //server/external -U dmitry%password -c ls
> Domain=[WORKGROUP] OS=[Windows 6.1] Server=[Samba 4.5.12-Debian]
>   .                                   D        0  Mon Dec 25 22:39:29 2017
>   ..                                  D        0  Sun Apr 24 19:50:32 2016
>                 1922727728 blocks of size 1024. 45707296 blocks available
> 
> How can I change the behaviour so that I/O error is appropriately reported,
e.g.
> the client gets "Error reading directory"?
> 
> Many thanks.
We'll need to see the log from smbd to see where the IO error is
getting dropped.

On 2018-01-02 19:46, Jeremy Allison wrote:
> We'll need to see the log from smbd to see where the IO error is
> getting dropped.
 
Thanks for reply. Which log level should be sufficient?
My current setting is

log level = 2

and I don't see any error(s) in /var/log/samba/log.smbd


-- 
With best regards,
Dmitry